PROYECTO COMUNIDAD EL SARDINAL 1, 2 MUNICIPIO DE JINOTEGA DEPARTAMENTO DE JINOTEGA ESTRUCTURAS PRIMARIAS Y SECUNDARIAS EN POSTES DE CONCRETO Y MADERA 14.4/24.9 K

General discussion

In the current thesis, I investigated how the knowledge of conspecifics, living and non- living entities is represented in the brain. Results showed that, naming pictures of social groups was selectively impaired in patients both with focal lesions (brain tumors) and diffuse brain atrophy (neurodegenerative diseases) in the left hemisphere. Overall, this set of areas includes insula, inferior frontal gyrus (IFG), amygdala (only in Chapter 2 in tumor patients), and medial prefrontal cortex (MPFC) (only Chapter 3 in dementia patients), areas often found to be associated with affective processing (for review, see Pessoa, 2008). Moreover, a further investigation on healthy participants revealed that the stimulation of a specific area within the opercular part of IFG (IFGop), which was found associated with conspecifics knowledge in the meta-analysis and in both the neuropsychological studies in this thesis, led to slower response time when the valence of the stimuli was negative. In this section I will discuss how these findings, including the anatomical lateralization, can be accommodated within current models on person-specific recognition and general semantic knowledge.

The investigation of conspecific knowledge has for long coincided with the study of semantic memory for familiar and famous persons. In the light of these studies, Bruce and Young (1986) proposed a model, revised by Burton et al. (1990, 1999) and, more recently, by Blank et al. (2014), to explain how we recognize individuals from faces, voices and names. The model proposed that distinct recognition units would encode specific-persons’ face, voice and name representations, and that each recognition unit would be connected to a person identity

node (PIN), involved in the multimodal representation of each individual. The different versions of the model do not agree on how person-related information (e.g., the social group they belong to) is recollected. On the one hand, each PIN is relevant in forming a multi-modal semantic representation of the other individuals, including person-related semantic

information (Bruce and Young, 1986), while on the other, they would be connected to the semantic information unit, representing only a merging zone between the different perceptual representation (Burton et al., 1990; 1999).

The meta-analysis I reported in Chapter 1 of my thesis revealed that the brain network was selectively associated with conspecific representation that included several areas in both hemispheres. More specifically the network included visual areas (bilateral occipital cortices, bilateral fusiform gyri, bilateral posterior STS) possibly corresponding to face recognition units (FRUs), auditory areas as the right STS possibly corresponding to voice recognition units (VRUs) and areas often found to be associated with emotional processing such as the

amygdala bilaterally, medial prefrontal cortex, and left inferior frontal gyrus. Interestingly, no specific area within the anterior temporal lobes (ATLs) (PINs) has been found to be

associated with the processing of conspecifics. The lack of an involvement of areas within the anterior temporal lobes can be explained in two ways. First, this might be due to the

deformation of the magnetic field induced by air-filled structures close to ATLs, that, without distortion-corrected fMRI protocols, would induce a loss of signal in those areas (Murphy et al., 2007). Second, the studies included in the meta-analysis I performed might have used tasks and stimuli (e.g., unfamiliar faces) that failed to tap the identity of each individual.

The finding of unimodal processing areas (i.e., visual, auditory and emotion) associated with conspecific knowledge can be interpreted within the context of the model by Bruce and Young (1986) on person-specific recognition at least in two different ways. First, this

processing could be interpreted as recognition unit coded in a different modality (i.e., a putative emotion recognition unit), possibly involved in representing the attitude towards each individual, and second, it might represent the outcome of the recollection of person- related knowledge. Hence, the two explanations entangle two different roles of affective processing in conspecifics knowledge: According to the former account, affective processing is part of the conspecifics conceptual knowledge, while according to the latter it would merely be an epiphenomenon of conspecific knowledge recollection. Moreover, the prediction stemming from the two accounts is that, for the former, the lesion of emotional processing area would lead to the impoverishment of the conceptual knowledge about conspecifics, with spared abilities of recognizing people from faces, voices and names, while for the latter it would lead to no conceptual deficit of conspecifics knowledge.

The findings reported on patients with brain tumors and on patients with dementia reported respectively in Chapter 2 and Chapter 3,in which I investigated semantic knowledge about social groups and other entities (i.e., living things and non-living things), showed that the lesion of areas often associated with affective processing led to a selective impairment in naming that can be attributed to a degradation of lexical-semantic knowledge about social groups (i.e., basic level knowledge about conspecifics). Moreover, in Chapter 4 healthy

participants performed a lexical decision task with nouns of social groups and exemplars from other categories and I observed that participants were slower RT when transcranial magnetic stimulation (TMS) was applied on one of these areas, i.e. IFGop, than in the control condition (i.e., vertex) specifically when the valence of the presented word was negative. This latter finding suggests that IFGop is sensitive to the valence of the stimuli. Taken together the findings reported in my thesis strongly suggest that affective features are relevant in building up a conceptual representation of our conspecifics. They also lead me to reject the hypothesis that the activation of emotion-related areas might be an epiphenomenon of the conspecifics

perception as suggested by the meta-analysis on neuroimaging studies about conspecifics (Chapter 1).

The model of person recognition by Bruce and Young (1986) does not directly take into account the different levels of specificity of the human stimuli that might contribute towards their representation. A human stimulus might be recognized as a specific individual, say a person we personally know (unique entities), but also as belonging to a category of individuals, that is a social group (supra-ordinate level). Unique entities and supra-ordinate human stimuli might differ in several respects: unique entities are usually associated with specific semantic knowledge, they are usually associated also with episodic memory, they are denoted by a proper name, they might be associated with richer perceptual knowledge than supra-ordinate knowledge (which tends to be prototypical) (Ross and Olson, 2012; Wang et al., 2016) and the distinction among unique entity elements might be based on more subtle features than for supra-ordinate (Gorno Tempini and Price, 2001). The investigation on unique entities revealed that, although different categories of entities (e.g., famous persons and famous buildings) tend to be associated with distinct sites of activations, all unique entities were associated with the function of the temporal poles (Gorno Tempini and Price, 2001; Ross and Olson, 2012). Coherent with these findings some fMRI experiments showed that ATL tend to be activated when the task requires a semantic discrimination between sub- ordinate entities, but not between supra-ordinate entities (Tyler et al., 2004; Rogers et al., 2006). Moreover, patients with semantic dementia - a neurodegenerative disease associated with the atrophy of ATL - tend to show semantic impairments, which are more severe for subordinate knowledge than for supra-ordinate concepts (Patterson et al., 2006; Rogers et al., 2015). This pattern of findings suggests that along the temporal lobes, there might be a

gradient of representation, from posterior to anterior, processing, respectively, from general to specific concepts (Martin, 2007). However, other studies made the picture more complex.

Indeed, ATL was found to be involved also in basic/supra-ordinate level of processing in TMS (Pobric et al., 2010) and FMRI (Visser et al., 2011) studies. Moreover, results from

neuropsychological studies in Chapter 3, and to a smaller extent, in Chapter 2 of my thesis, showed that also social group knowledge, which codes information about individuals at a supra-ordinate level, was associated with temporal polar damage. Together my findings, strongly suggest that ATL might also be involved in the processing of basic-level concepts.

However, a critical aspect that must be considered when discussing about the

representation of conspecifics, as well as about other categories of entities, is the functional lateralization. Although the meta-analysis (Chapter 1 of this thesis) highlighted a pattern of activations associated with conspecifics processing (except for the left IFG) mostly bilaterally, patients’ results (reported in Chapters 2 and 3 of this thesis) highlighted the involvement of a left-lateralized network of areas. This pattern of findings might be explained, at least, in two ways. The lateralization might be explained by the type of task I used in the experiments, but it could also be specific for the representation of this category of knowledge. The relative contribution of left and right hemispheres to semantic representation is still a matter of debate. Indeed patients with a left-lateralized damage tend to have more impaired production abilities compared with their receptive abilities, while patients with right lateralized (or predominantly right–lateralized) damage tend to lead to both productive and receptive abilities or to very mild impairments (in case of unilateral brain damage) (Lambon Ralph et al., 2001; Rice et al., 2015). To explain these findings, two main theories were proposed. First, it has been proposed that semantic knowledge might be differently specialized into the two hemispheres, with the left ATL involved in representing verbal knowledge and right ATL representing non-verbal (e.g., visual, auditory) knowledge (Snowden et al., 2004; 2012). Second, another model proposed that semantic knowledge is represented bilaterally, but left- lateralized semantic storage would have a privileged access to phonological areas, leading

left-lateralized semantic deficits to be expressed mainly as lexical impairments. Both the neuropsychological studies reported in this thesis (Chapters 2 and 3) demonstrate that patients with unilateral and diffuse brain damage showed impaired productive abilities (picture naming task) but spared recognition abilities (word-to-picture matching task). Moreover patients’ picture naming deficits were mainly associated with lesions of a left- lateralized network. Hence, according to the model by Snowden (2004, 2012) the social group deficits we measured were basically at the lexical level, while according to the model by Lambon Ralph et al. (2001) they were at the level of the semantic representation. Although this pattern of findings is consistent with both theories on semantic deficits lateralization, right lateralized damage to the very same areas in patients with brain tumors did not give rise to any semantic deficit (on either production or recognition), thus the model by Snowden et al. (2004; 2012) cannot accommodate my results.

A critical issue about the person recognition model is how the specific processing of conspecifics relates to other categories of knowledge. For instance, several studies suggested that animate and inanimate entities tend to be processed in discrete neural areas, such as the inferior part of the ATL for animate entities, and the inferior parietal lobe (IPL)/posterior temporal lobe for inanimate entities (Gainotti, 2000). This pattern of findings was

consistently found using different methodologies (neuropsychology, fMRI, TMS,

electrophysiology) and led to propose that these two categories might be processed by different neural systems (see also Brambati et al., 2006; Anzellotti et al., 2011; Damasio et al., 1996; Devlin et al., 2002, Chan et al., 2011) and that, while animate entity processing might rely on areas within the ventral visual pathway (inferior temporal cortex) (Kravitz et al., 2013), inanimate entities might rely on areas within the dorsal visual pathway (IPL) (Kravitz et al., 2011). In this context the double dissociations in patients between the knowledge about conspecifics at both unique-entity (Hanley et al., 1989; Evans et al., 1995; Miceli et al., 2000;

Kay and Hanley, 2002; Lyons et al., 2006; Thompson et al., 2004) and social group (Rumiati et al., 2014) level of specificity, and living or non-living entities, suggest that conspecifics are processed independently of the exemplars from the other two categories.

My thesis adds a further piece of information about the way in which conspecifics are represented in the brain. In fact I have found evidence that the processing of conspecifics at a basic level of specificity is associated with areas typically involved in processing emotions. This might be interpreted in different ways depending on the model. For instance, according to the domain-specific hypothesis (DSH, Caramazza and Shelton, 1998; Mahon and

Caramazza, 2011) distinct domains of entities such as animals, plants, artifacts and conspecifics, are processed by different neural networks as a result of the evolutionary

pressure. Thus, the network of areas that I found to be selectively associated with social group processing, seems to support the view of a domain-specific representation for conspecifics. This interpretation is confirmed by the findings of the meta-analysis reported in Chapter 1 in which the same areas (except for insula) were found to be activated in studies which directly contrasted social vs. non-social (e.g., animals, places, objects) stimuli. However, the TMS study reported in Chapter 4 of the current thesis clarified that one if these areas, the left IFGop rather than being involved in the categorical processing of social groups, it might more likely be involved in processing emotional stimuli.

Conversely, other theories highlighted the role of modality-processing areas in representing concepts. According to the sensory-functional theory (SFT) (Warrington and Shallice, 1984), for instance, our cognitive system processes the distinct conceptual features into parallel modality-dependent semantic systems (i.e., sensory, motor/functional). Thus, categorical deficits may emerge as a consequence of damage to one of these subsystems, since the representation of distinct categories relies on different semantic features.

My findings can be better accounted for if, beyond the classical formulation of the theory, holding a distinction between sensory (mainly visual) and functional features, we also consider the emotional modality dedicated to process the emotional features of conspecifics. Indeed, in my study involving patients with neurodegenerative diseases (Chapter 3), the deficit in picture naming was associated with the atrophy at the level of bilateral inferior temporal cortices for natural categories (the ending point of the ventral visual pathway (Kravitz et al., 2013)), with dorso-lateral prefrontal cortex for artificial categories (relevant in motor processing (Hetù et al., 2013)), and emotion-related areas for social groups.

Moreover, Simmons and Barsalou (2003) proposed a hierarchical model, the

conceptual topography theory (CTT), in which exemplars’ features are represented within modality-specific areas that are integrated, through the sequential processing within distinct convergence zones (CZs), in order to assemble categorical representations. While analytic and holistic CZs are relevant in order to integrate features sharing similar properties, modality and cross-modal CZs are relevant for integrating conceptual properties in order to implicitly form categorical knowledge. Moreover, they proposed that, within convergence zones, neurons representing similar features are located topographically close. This complex model has the potential for explaining many different patterns of patients’ deficits, ranging from isolated categorical deficits, to multi-categorical deficits (i.e., deficits in the knowledge of more than one category, with some categories preserved), to global semantic impairments.

Categorical semantic deficits might emerge as a consequence of the damage modality and cross-modal CZs (respectively on uni-modal and multi-modal concepts), or in the case of analytic and holistic CZs lesion, only if the category relies on the damaged property. Thus the association between left insula, left anterior cingulate/medial PFC and left IFG lesion with the lexical-semantic impairment in patients’ naming social groups is not easy to be explained in the light of the CTT. The model easily explains the categorical knowledge deficits reported in

the study, but we are not able to say whether these areas are involved in the representation of some specific properties of social stimuli (which could be associated with the lesion of

analytic or holistic CZs) or of their uni-modal categorical knowledge (modality CZs for the emotional modality). However, the posterior part of the IFG (pars opercularis) has been proposed as a hub connecting affective and motor networks (Jezzini et al., 2015; Gerbella et al., 2016; Morecraft et al., 2012), and might be interpreted, in the light of CTT, as a cross- modal CZ. Moreover, the pattern of findings on natural and artificial stimuli, which coincided with the ending points of, respectively the ventral and the dorsal visual pathways, could also be interpreted according to the CTT, as modality CZs (or alternatively, cross-modal CZs). Hence, the CTT, differently from the SFT, is able to explain why, the deficit in naming a certain category of stimuli might be associated with the atrophy within different modality processing areas, when is associated with other categories.

To sum up, the findings of the current thesis, suggest that conceptual knowledge is associated with modality-specific processing areas, and specifically that social group

representation might interact with emotional features. From a theoretical point of view this is more easily explainable by the CTT, although is not possible to exclude the role of domain- specific networks explaining my findings, as suggested by the DSH. However, further research is necessary, in order to better understand how we represent our conspecifics and the specific role of each emotion-related area associated with social group lexical-semantic knowledge.

References

Adolphs, R., Tranel, D., Damasio, H., & Damasio, A. (1994). Impaired recognition of emotion in facial expressions following bilateral damage to the human amygdala. Nature, 372(6507), 669-672.

Adolphs, R., Damasio, H., Tranel, D., Cooper, G., & Damasio, A. R. (2000). A role for somatosensory cortices in the visual recognition of emotion as revealed by three-dimensional lesion mapping. The Journal of Neuroscience, 20(7), 2683-2690.

Adolphs, R. (2009). The social brain: neural basis of social knowledge. Annual review of psychology, 60, 693.

Albanese, E., Capitani, E., Barbarotto, R., & Laiacona, M. (2000). Semantic category dissociations, familiarity and gender. Cortex, 36(5), 733-746.

Amodio, D. M., & Frith, C. D. (2006). Meeting of minds: the medial frontal cortex and social cognition. Nature Reviews Neuroscience, 7(4), 268-277.

Amodio, D. M. (2014). The neuroscience of prejudice and stereotyping. Nature Reviews Neuroscience, 15(10), 670-682.

Andics, A., McQueen, J. M., Petersson, K. M., Gál, V., Rudas, G., & Vidnyánszky, Z. (2010). Neural mechanisms for voice recognition. Neuroimage, 52(4), 1528-1540.

Anzellotti, S., Mahon, B. Z., Schwarzbach, J., & Caramazza, A. (2011). Differential activity for animals and manipulable objects in the anterior temporal lobes. Journal of Cognitive

Neuroscience, 23(8), 2059-2067.

Appollonio. I.. Leone. M.. Isella. V.. Piamarta. F.. Consoli. T.. Villa. M. L.. ... & Nichelli. P. (2005). The Frontal Assessment Battery (FAB): normative values in an Italian population sample. Neurological Sciences. 26(2). 108-116.

Aron, A. R., Fletcher, P. C., Bullmore, E. T., Sahakian, B. J., & Robbins, T. W. (2003). Stop- signal inhibition disrupted by damage to right inferior frontal gyrus in humans. Nature neuroscience, 6(2), 115-116.

Ashburner, J., & Friston, K. J. (2000). Voxel-based morphometry—the methods. Neuroimage, 11(6), 805-821.

Ashburner. J. (2007). A fast diffeomorphic image registration algorithm. Neuroimage. 38(1). 95-113.

Baillargeon, R. (1993). The object concept revisited: New directions in the

investigation of infants’ physical knowledge. Visual perception and cognition in infancy, 23, 265-315.

Baird, A. D., Scheffer, I. E., & Wilson, S. J. (2011). Mirror neuron system involvement in empathy: a critical look at the evidence. Social neuroscience, 6(4), 327-335.

Barbarotto, R., Laiacona, M., Macchi, V., & Capitani, E. (2002). Picture reality decision, semantic categories and gender: A new set of pictures, with norms and an experimental study. Neuropsychologia, 40(10), 1637-1653.

Barsalou, L. W., Simmons, W. K., Barbey, A. K., & Wilson, C. D. (2003). Grounding

conceptual knowledge in modality-specific systems. Trends in cognitive sciences, 7(2), 84-91. doi:10.1016/S1364-6613(02)00029-3

Bates, E., Wilson, S. M., Saygin, A. P., Dick, F., Sereno, M. I., Knight, R. T., & Dronkers, N. F. (2003). Voxel-based lesion–symptom mapping. Nature neuroscience, 6 (5), 448-450. doi:10.1038/nn1050

Beauchamp, M. S., Lee, K. E., Haxby, J. V., & Martin, A. (2003). FMRI responses to video